Spatial Autocorrelation

Spatial autocorrelation is a metric that measures the relationship of a variable in a particular location with the same variable in other locations.

In practice, it is commonly calculated using neighboring observations to identify if the location influences the variable’s value.

A positive spatial autocorrelation of a particular variable indicates the similarity of that variable among neighboring observations, meaning that similar values tend to be together. For instance, you can consider house prices as an example because the location influences a house price, causing neighboring houses to have similar prices.

A negative spatial autocorrelation indicates that neighboring observations have dissimilar values, causing a checkerboard pattern or the presence of spatial variance across the region. This is less common in social phenomena. One example is the distribution of supermarkets of different brands, or of hospitals. To avoid direct competition, the distribution of the supermarks or hospitals should be away from each other to have a better spatial coverage. In other words, it follows a pattern of negative spatial dependence.